The Scale of the Problem
Every day, the global oil and gas industry burns off enough natural gas to power entire cities. This is not a technical failure. It is a structural one: gas produced alongside oil at remote wellsites has nowhere to go when pipelines are full, absent, or economically unjustifiable to build.
In 2024, global gas flaring reached 151 billion cubic meters (bcm) 1, the highest level since 2007 according to the World Bank's 2025 Global Gas Flaring Tracker Report. That represents a 2% increase over 148 bcm in 2023 and part of a cumulative 9% rise since 2022. Flaring intensity also climbed to 5.1 cubic meters of gas wasted per barrel of oil produced 2, according to the Environmental Defense Fund's 2025 Flaring Update.
The United States ranked fourth globally in flaring volume in 2024, having risen from sixth place in prior years. Russia, Iran, and Iraq led the list, but the top nine countries together account for roughly 75% of all global flaring 3 while representing less than 50% of global oil production.
The economic cost is staggering. At EU import prices, the World Bank estimated $48 billion in gas value was destroyed by flaring in 2023 4. A Foreign Policy Association analysis 5 placed the World Bank's own figure at $50 billion wasted annually. FlareIntel's 2025 outlook 6 puts the annual loss at $30 billion, treating 150 bcm as roughly 4% of global gas production. Either figure represents a capital destruction problem of the first order.
To achieve zero routine flaring by 2030, the global industry would need to reduce flaring by 23% per year from current levels, a target the EDF now regards as out of reach 7. The infrastructure investments required, combined with the pace of new drilling, make the problem structural rather than transitional.
The Williston Basin: Ground Zero
North Dakota's Williston Basin is the most consequential single geography in the US flaring picture. In 2024, the state vented and flared approximately 67,700 million cubic feet (MMcf) of natural gas 8, accounting for roughly one-fifth of total US gas vented and flared in that year, according to Arbo's analysis of EIA data.
On a daily basis, NDIC Director's Cut reports 9 show North Dakota flaring between 160 and 175 million cubic feet per day (MMCFD) throughout 2024 and into 2025. That translates to roughly 60 Bcf per year, even as the state's overall gas capture rate sits at 95%.
The 95% capture figure is genuinely impressive compared to the 36% flaring rate North Dakota recorded in September 2011 10. But at 3.4 billion cubic feet per day of total gas production, the remaining 5% represents a massive and growing absolute volume.
The underlying driver is a structural shift in reservoir behavior. The Bakken's gas-to-oil ratio (GOR) has tripled in a decade, from roughly 1 MCF per barrel in 2014 to approximately 3 MCF per barrel in 2023 11, per Novi Labs analysis. With North Dakota producing around 1.2 million barrels of oil per day, the associated gas volume now exceeds 3.4 Bcfd and continues to grow. Wells simply get gassier as reservoir pressure declines. This is not an anomaly that infrastructure build-out will fully resolve. It is a permanent feature of late-stage unconventional production.
The North Dakota Oil & Gas Association's 2024 Production Report 12 confirms the production base: 437.8 million barrels of oil and 1.253 trillion cubic feet of total gas produced from approximately 19,266 producing wells. Roughly 3,200 of those wells 13 experience active or intermittent flaring, including wells connected to gathering networks that still flare due to pressure differential mismatches between the wellsite separator and the gathering system pressure.
Pipeline infrastructure is the root cause of the persistent gap. The EIA described the structural challenge as early as 2011 14: gathering pipelines, processing plants, and transportation pipelines have consistently failed to expand at the same pace as oil production, effectively stranding the associated natural gas. That dynamic persists. The EDF's 2025 update specifically cited the Bakken 15 as a driver of the US flaring increase in 2024, noting that produced gas volumes exceeded takeaway capacity.
Large-scale infrastructure projects are underway but will not solve the last-mile problem. The Intensity Infrastructure/Rainbow Energy Center pipeline 16, announced in June 2025, will add 1.1 Bcfd of capacity via a 136-mile, 36-inch line from Watford City to Washburn, targeting service in early 2029. Basin Electric's Pioneer Generation Station Phase IV 17 added 580 MW of gas-fired generation northwest of Williston, completed in October 2025. A 1,490 MW Bison Generation Station 18 in Williams County is slated for 2029 to 2030. These projects are significant. They consume pipeline gas, not flare gas. They do not address the 160 to 175 MMCFD of gas flared daily at remote, small-volume, or newly drilled sites that pipelines cannot economically reach.
The Economics of Stranded Gas
The economics of stranded flare gas as an energy input are unlike anything available in conventional power markets. The gas is effectively free, or close to it.
At the wellsite, gas that would otherwise be flared carries near-zero or negative economic value. Operators must either burn it, pay to dispose of it, or install equipment to use it. This creates an acquisition cost of 10 to 30 cents per thousand cubic feet (MCF) 19, against a Henry Hub market price consistently above $3 per MCF. That 10x to 30x discount on fuel input is the foundational advantage of the wellsite generation model.
When converted to electricity using modern reciprocating gas generator sets, stranded gas produces power at a fully loaded cost of under $0.01 per kilowatt-hour 20 (under $10 per megawatt-hour). Compare this to industrial grid rates in North Dakota of $0.07 to $0.09 per kilowatt-hour, or US commercial electricity averaging $0.10 to $0.13 per kilowatt-hour. The delivered cost advantage is 7x to 13x.
This advantage even exceeds the best available grid-tied generation options. The EIA's Annual Energy Outlook 2025 LCOE report 21 prices new natural gas combined-cycle plants at $37.58 per MWh, gas combustion turbines at $48.78 per MWh, onshore wind at $29.58 per MWh, and utility-scale solar at $26.06 per MWh (both renewable figures include production tax credits). Stranded gas generation at under $10 per MWh undercuts every one of these by 3x to 5x, without requiring grid interconnection, transmission infrastructure, or battery storage.
The capital structure of a modular wellsite generation unit reflects the remote deployment requirements. Gas input requirements scale linearly 22: a 1 MW unit requires approximately 264 Mcf per day of gas input, meaning a site flaring 1 MMscfd (1,000 Mcf/day) can support roughly 130 to 170 kW of continuous generation using modern reciprocating engine technology, per the US Department of Energy's NETL Stranded Gas Roadmap.
The project economics at the wellsite are compelling even before compute monetization. A 1 MW unit with a fully installed capital cost of $700,000 to $1.2 million, running at $0.08 per kilowatt-hour equivalent electricity value across 8,000 annual operating hours, generates roughly $640,000 in annual power value at zero fuel cost. Simple payback is one to two years.
The compute premium changes the model entirely. Deploying GPU compute infrastructure directly at the wellsite allows operators to monetize electricity not at grid rates but at AI compute rates. H100 GPU cloud pricing runs $8 to $12 per GPU-hour 23 on distributed compute platforms, versus $25 to $30 on AWS or Azure. At the unit economics of a fully deployed 2 MW module, gross cash flow of $400,000 to $500,000 per month per site 24 is achievable, compressing payback periods to 18 to 24 months even on fully installed capital costs of $7 to $8 million per module.
FlareIntel's ESG investment analysis 25 confirms this directionally: flare capture projects at sites with 0.5 MMscfd or more of consistent flow can break even in under two years and typically generate negative marginal abatement costs of up to -$40 per tonne of CO2 equivalent. This means capturing flared gas is not just environmentally beneficial; it is cash-flow positive by the metric used to evaluate carbon reduction investments.
From Waste Stream to Compute Asset
The technical model for converting flare gas into compute power is straightforward, though deploying it at the wellsite involves solving a series of practical engineering and logistical challenges.
The core unit is a skid-mounted or containerized reciprocating gas engine generator, packaged with all necessary balance-of-plant equipment: gas conditioning, control systems, electrical distribution, and thermal management. The generator connects directly to the wellsite separator or gathering header, consuming gas that would otherwise go to the flare stack. Output is low-voltage power, distributed on-site to computing equipment housed in adjacent containerized server modules.
Because the entire system is modular, it can be assembled off-site and transported to the wellsite in sections on standard flatbed trailers. This is the key logistical advantage over conventional power infrastructure, which requires civil works, electrical transmission, and utility interconnection processes that often take years.
Giga Energy reports deployment timelines of 6 to 8 months 26 for AI-ready wellsite data centers using pod-based modular construction, compared to an industry standard of 24 to 36 months for conventional data centers. That speed advantage is operationally significant because wellsite flaring volumes can shift as reservoir pressure declines or new wells come online. A fast-deploy, fast-relocate capability is not just convenient; it is a prerequisite for operating at scale across a distributed set of small and mid-sized flare sites.
The gas conditioning requirements depend on the wellsite gas composition. Bakken associated gas is relatively rich in heavier hydrocarbons (ethane, propane, butane), which increases the heating value but may require knock-out separators to prevent liquid carryover into the engine. The NETL Stranded Gas Roadmap 27 identifies gas composition variability and pressure fluctuations as the primary technical challenges for wellsite generation at small scales. Modern lean-burn reciprocating engines with real-time air-fuel ratio control handle these variations effectively, making the technology commercially mature.
Power management at the computing module level requires active load-following or curtailment capability. When wellsite gas pressure drops due to reservoir depletion, scheduled maintenance, or operator intervention, the computing load must reduce commensurately. This is less problematic for cryptocurrency mining workloads (where compute is entirely interruptible) but requires careful load scheduling for AI inference workloads where latency and availability SLAs must be managed. The sector's early operators have handled this through a combination of load prediction algorithms and contractual structures that price the power variability into compute pricing.
The Environmental Case
Open-air gas flaring looks simple: burn the gas, produce CO2, avoid the worse outcome of methane venting. The reality is more complicated, and the gap between regulatory assumptions and measured performance is where the environmental case for controlled combustion becomes decisive.
Regulators typically assume flares operate at 98% methane destruction efficiency. Empirical studies paint a different picture. EDF's research found actual average combustion efficiency of approximately 91% 28, meaning roughly 9% of methane in flare gas is emitted un-combusted. Methane carries 80 times the warming potential of CO2 over a 20-year timeframe 29, according to IPCC assessments. The result: actual greenhouse gas impact from flaring is 2x to 5x higher than CO2-only metrics suggest.
FlareIntel's ESG analysis 30 calculated total flaring emissions, accounting for methane slip, at 1.2 billion tonnes CO2-equivalent per year globally. The World Bank's GFMR Partnership 31 estimated 389 million tonnes of CO2 equivalent from 2024 flaring, of which 46 million tonnes was un-combusted methane.
For North Dakota specifically: approximately 60 Bcf flared annually, at standard combustion, produces roughly 3.3 million metric tons of CO2 per year. Adding the methane slip at 9% un-combusted, the total rises to an estimated 8 to 9 million metric tons of CO2-equivalent annually from North Dakota flaring alone.
Modern gas engine generators running on the same wellsite gas achieve 99%+ combustion efficiency 32, compared to the 91% empirical average for open flares. MARA's wellhead data center program reported combustion efficiency of up to 99% 33 against a typical flare efficiency of approximately 92%. The measurable difference is a roughly 63% reduction in CO2-equivalent emissions per unit of gas consumed, along with complete conversion of chemical energy to useful electrical work rather than atmospheric heat.
The IEA estimates the oil and gas sector could reduce methane emissions by 75% using existing technology 34, with up to 40% of those reductions achievable at no net cost because of the value of recovered gas. Wellsite power generation is precisely the type of project the IEA identifies as cost-neutral or cost-positive: the captured gas pays for the equipment that captures it.
Carbon credit monetization provides an additional revenue layer, though not yet the primary economic driver. Flare reduction projects can generate verified methane avoidance credits on registries such as Verra. Annual methane credit retirements have tripled to more than 18 million metric tonnes since 2019 35, indicating growing buyer demand. Higher-quality credits, including those from methane avoidance projects with rigorous measurement protocols, now trade at a 30% premium 36 over lower-tier credits. Major data center operators including Microsoft, Google, Amazon, and Meta are active buyers in voluntary carbon markets, creating a natural demand base for the credits that wellsite generation projects can produce.
The Regulatory Landscape
Regulation creates both the floor and the ceiling for the stranded gas-to-compute market. The floor is North Dakota's mandatory gas capture regime. The ceiling was a proposed federal methane fee that has since been removed.
The North Dakota Industrial Commission (NDIC) established a statewide 91% gas capture requirement 37 effective November 2020, the culmination of a progressive tightening that began in 2014 when North Dakota mandated 74% capture. Operators failing to meet the 91% threshold at the field level can be denied new drilling permits. The regulation explicitly credits on-site beneficial use of captured gas, including power generation, toward the capture requirement. This is a direct regulatory incentive for wellsite generation: operators can satisfy capture mandates by burning gas productively rather than flaring it.
North Dakota banned venting of natural gas entirely in 2014, requiring all unprocessed gas to be either captured or flared rather than released directly to atmosphere. Some legacy and remote fields still record low capture rates: Zahl at 50%, Fort Buford at 48%, Bar Butte at 58%, and Briar Creek at 54%, per NDIC data. These fields represent precisely the target market for distributed modular generation: remote, structurally underserved by infrastructure, and facing regulatory pressure to improve.
At the federal level, the picture has shifted significantly under the current administration. The EPA Inflation Reduction Act Waste Emissions Charge, which would have imposed $900 per metric ton of methane on large emitters in 2024, rising to $1,500 per metric ton by 2026, was repealed. President Trump signed a Congressional Review Act resolution on March 14, 2025 38 disapproving the rule. The EPA removed the charge from the Code of Federal Regulations entirely by May 2025. The financial penalty driver is gone under current federal policy.
The EPA's 2024 New Source Performance Standards (NSPS OOOOb/c), which required routine flaring elimination, have also been significantly delayed. An EPA Interim Final Rule in July 2025 39 extended compliance deadlines by 18 months, with a November 2025 Final Rule 40 pushing reporting compliance to November 2026. The EPA is separately reconsidering the underlying 2024 methane requirements as part of a broader deregulatory agenda.
This regulatory context matters for demand forecasting. Federal enforcement pressure is lighter in 2026 than it would have been under the prior administration. But North Dakota state regulations remain intact and represent a more durable regulatory floor than federal rules, which change with administrations.
On the positive regulatory side, the FLARE Act 41, introduced by Sen. Ted Cruz in March 2025, would provide permanent 100% bonus depreciation for flaring and venting mitigation systems, explicitly including equipment that converts captured gas to electricity, computational power, or digital asset mining. If enacted, it would materially improve after-tax returns on wellsite generation equipment. The bill aligns with the current administration's energy production and AI infrastructure priorities and has backing from the Digital Power Network.
The proposed GRID Act 42, introduced bipartisan in the Senate in February 2026, would require new data centers larger than 20 MW to source power from off-grid generation rather than the electrical grid, with a 10-year phase-in for existing facilities. If passed, it would be among the most significant regulatory tailwinds ever created for the stranded gas-to-compute model. Whether or not it passes, the fact of its introduction reflects legislative awareness that grid-dependent data center growth is creating systemic electricity infrastructure stress.
Market Validation
The stranded gas-to-compute model has been commercially validated at scale. The question is no longer whether it works. The question is who builds the next generation of it.
Crusoe Energy: The Pioneer's Pivot
Crusoe Energy Systems, founded in Denver in 2018, created the Digital Flare Mitigation (DFM) category and proved the model at industrial scale. By 2024, Crusoe had converted more than 10.4 billion cubic feet of flare gas into approximately 1.3 terawatt-hours of electricity 43, supplying 87% of its total power needs and avoiding over 1.3 million metric tonnes of CO2-equivalent, equivalent to removing 300,000 cars from the road for a year. Cumulative since 2018: more than 21 billion cubic feet of flaring avoided and approximately 2.7 million metric tonnes of CO2-equivalent avoided.
On March 25, 2025, Crusoe announced the divestiture of its entire bitcoin mining operation and Digital Flare Mitigation business to NYDIG 44, a $7 billion digital asset firm and subsidiary of Stone Ridge Holdings. The transaction transferred 425+ modular data centers across seven US states and Argentina, representing more than 270 MW of power generation capacity, along with approximately 135 Crusoe employees. Crusoe retained a significant equity stake in the resulting joint venture. Terms were not disclosed; Crusoe carried a $2.8 billion valuation at the time. CNBC reported the strategic rationale 45: Crusoe chose to focus entirely on its hyperscale AI data center business, including a 1.2 GW campus in Abilene, Texas, built for Project Stargate, and a pipeline of more than 10 GW in development.
NYDIG, operating under Stone Ridge's 10 GW natural gas portfolio strategy, now operates the DFM business. The technology was not retired. It was acquired because it generates value. The pioneer of the model chose hyperscale AI infrastructure over distributed wellsite operations, creating a clear market vacancy in the distributed segment.
Giga Energy: Distributed at Scale
Giga Energy, a Texas-based firm co-founded by Brent Whitehead and Matt Lohstroh, has deployed more than 150 MW of containerized modular data centers 46 across sites in Texas and internationally, with containers holding more than 300 ASIC miners or GPUs per unit. The company completed an international expansion in March 2024 into Argentina's Mendoza Province 47, utilizing flare gas from the Vaca Muerta shale formation, home to the world's second-largest shale gas reserve. Domestically, Giga operates a joint venture with Atlas Power to develop stranded and flare gas sites in the Williston Basin 48, directly applying the distributed model to the North Dakota Bakken.
MARA Holdings: 25 MW Across the Bakken
Marathon Digital Holdings (MARA) partnered with NGON (Natural Gas On-site), an oilfield gas mitigation provider, to build a 25 MW micro data center program distributed across wellheads in Texas and North Dakota 49. Sites began energizing in September 2024 and reached full operation in January 2025. In the first five months, the program reduced 29,300 metric tonnes of CO2-equivalent 50, equivalent to removing 6,800 gasoline vehicles from the road, while consuming approximately 5,000 MCF of associated gas per day. MARA describes these operations as delivering the lowest energy cost per bitcoin across its 1.7 GW global fleet.
ExxonMobil: Oil Major Entry via Low-Carbon Data Centers
ExxonMobil entered the picture from the supply side. The company is in advanced talks with power providers and technology firms 51 to build natural gas-fired data center power plants paired with carbon capture systems targeting 90% CO2 capture. A Final Investment Decision on the first project is targeted for late 2026. ExxonMobil's entry signals that even the largest integrated oil companies view gas-powered compute infrastructure as a strategic line of business, not a fringe experiment.
Together, these market entrants validate three points: the technology works at commercial scale, the economics are durable, and the opportunity is large enough to attract capital from bitcoin miners, AI infrastructure developers, and major oil companies simultaneously.
The Opportunity Ahead
The addressable inventory of stranded flare gas sites in the Williston Basin alone is substantial and well-defined.
Of North Dakota's approximately 19,266 producing wells, an estimated 3,200+ have active or intermittent flaring events 52, representing 17% of the total well count. These sites range from large central facilities flaring several MMscfd to small wellhead flares venting a few hundred Mcf per day. The economically viable subset, sites with consistent flare volumes of 500 Mcfd or more that can support modular generation units with acceptable utilization, numbers in the hundreds.
Research from the wellsite generation sector identifies approximately 468 qualifying sites in the Williston Basin 53 as viable candidates for distributed power generation deployment, representing roughly 125 MW of accessible power generation capacity 54 when aggregated. This is not a future projection. These are existing flare sites producing gas today.
The dynamics that created this inventory are getting stronger, not weaker. GOR tripling since 2014 means associated gas volumes continue to grow even without new drilling. Novi Labs identifies more than 38,000 proven and unproven drilling locations 55 in the Williston Basin at current pace, implying decades of continued production growth and continued gas stranding in remote areas. Pipeline infrastructure will continue to lag oil production in remote McKenzie and Dunn counties because the economics of rural pipeline construction for small-volume gas flows have never been favorable and will not become so.
The macro backdrop amplifies the opportunity. Data centers consumed 4% of total US electricity in 2024 56, a figure the IEA projects will grow to more than 6% by 2028 and potentially 11% by 2030. Goldman Sachs estimates 82 GW of new US generation capacity 57 is needed by 2030 to serve data center demand alone. Grid interconnection queues in saturated markets like Northern Virginia average 5 to 7 years, and GE Vernova's turbine backlog reached 80 GW in late 2025 58, with equipment sold out through 2029.
In that environment, a power source that can be deployed in 6 to 8 months without grid interconnection, at a delivered cost of under $0.01 per kilowatt-hour, using a fuel stream that would otherwise be wasted, is not a niche energy play. It is infrastructure positioned at the precise intersection of three structural trends: growing AI compute demand, grid interconnection scarcity, and an expanding inventory of stranded gas.
The flare gas power generation market 59 is projected to grow from $3.1 billion in 2024 to $6.3 billion by 2031, a 10.6% compound annual growth rate. North America accounts for roughly 40% of the global flare gas recovery system market. The model is not confined to the Williston Basin. The Permian Basin in West Texas, the DJ Basin in Colorado, the Anadarko in Oklahoma, and international formations including the Vaca Muerta in Argentina all present comparable or larger inventories of stranded flare gas.
The Williston Basin is where the regulatory structure, the operator base, the gas volumes, and the precedent set by early operators make the case most clearly. The gas is there. The demand for power is there. The technology to connect them has been proven at commercial scale. The window to build the distributed infrastructure layer is open now, before large-scale pipelines and centralized power plants absorb the gas volumes that modular generation can reach first.
Sources: World Bank 2025 Global Gas Flaring Tracker Report 60; EDF Flaring Update 2025 61; World Bank Global Flaring Data 62; FlareIntel 2025 Outlook 63; Foreign Policy Association 64; Arbo / North Dakota Gas Power Analysis 65; NDIC Director's Cut January 2025 66; Novi Labs North Dakota Update 67; NDOA 2024 Production Report 68; EIA Today in Energy 69; Intensity Infrastructure 70; Minot Daily News / Basin Electric 71; POWER Magazine / Bison Station 72; NETL Stranded Gas Roadmap 73; Dave Friedman Substack / Crusoe Economics 74; EIA AEO2025 LCOE Report 75; FlareIntel ESG Investment Analysis 76; Giga Energy Build Speed 77; EDF Flaring Update 2024 78; World Bank GFMR Methane 79; Dynamic Carbon Credits / IEA Methane 80; Crusoe 2024 Impact Report 81; MARA 25 MW Announcement 82; MARA Full Energization 83; Trellis / VCM 2025 84; Carbon Credits.com VCM 2025 85; EPA Waste Emissions Charge 86; EPA IFR July 2025 87; EPA Final Rule November 2025 88; Cruz FLARE Act 89; Troutman Pepper / GRID Act 90; Crusoe / NYDIG Divestiture 91; CNBC / Crusoe NYDIG 92; Atlas Power / Giga Energy JV 93; Giga Energy Argentina 94; ExxonMobil Data Center Plans 95; Goldman Sachs Data Center Power 96; Pew Research / Data Centers 97; Segue Infrastructure / GE Vernova 98; Cognitive Market Research / Flare Gas Market 99; Novi Labs Williston Basin 100